Video on Genvizu covers text-to-video, image-to-video and first/last-frame clips, plus canvas transitions, upscaling and reverse scripts. Most video models bill per second, so duration is the main cost lever: Grok Imagine Video 1.5 runs 2 credits per second at 480p, while Veo 3.1 bills a fixed 8-second clip. Write one shot per prompt, name the camera move, and check duration and ratio before generating.
Verified on 2026-09-03
| Field | Value |
|---|---|
| Typical inputs | A one-shot description, an optional first frame or reference image, the duration, the ratio and the resolution |
| Deliverables | Single clip, first/last-frame animation, canvas transition between two clips, upscaled master, written reverse script |
| Billing basis | Per second for most models; Veo 3.1 and Gemini Omni Flash 1.1 bill a fixed clip price |
| Duration range | Seedance 2.5 5-30s (5s minimum billing); Seedance 2.0 and MiniMax H3 4-15s; Grok Imagine Video 1.5 1-15s; Veo 3.1 fixed 8s |
| Cheapest per second | Grok Imagine Video 1.5 at 2 credits per second (480p) and 4 per second (720p) |
| Audio | MiniMax H3 outputs native stereo audio; check each model before assuming a soundtrack |
| Reference inputs | Seedance 2.0 family up to 9 reference images, 3 reference videos and 3 reference audio clips; Gemini Omni Flash 1.1 up to 7 reference images or a first/last frame pair |
| Acceptance | Camera move matches the brief, the subject holds together for the whole clip, no drifting limbs or morphing logos, cut points usable in an edit |
| Post-processing | Topaz upscale at 10 credits per second (1x/2x) or 18 (4x), billed on input duration, single file up to 50MB |
Prompt
Use the attached image as the first frame.
Slow 30-degree orbit around the product, camera at product height, constant speed.
The product stays centred and in focus for the whole shot; background stays softly out of focus.
Lighting unchanged from the first frame. No cuts, no zoom, no added text.
Duration 5s, 16:9.Starting from a still you already approved removes composition from the list of things that can go wrong, leaving only the motion to judge.
Prompt
A rain-soaked neon alley at night, seen from a low angle.
Camera: slow dolly forward, subtle handheld float, no cuts.
Subject: a lone figure in a dark coat walking away from camera, footsteps splashing shallow puddles.
Light: magenta and cyan signage reflecting in the water, wet asphalt highlights.
Mood: quiet, cinematic, no dialogue. Duration 6s, 16:9.One shot, one camera move, one action. Prompts that describe a whole scene sequence get compressed into a confusing single take.
Prompt
Vertical 9:16 clip for a social feed.
Static locked-off camera on a ceramic cup as steam rises in soft morning light.
Only the steam and a slight curtain movement animate; the cup and table are still.
Warm off-white palette, shallow depth of field, no text, no camera move.
Duration 5s.Naming the only elements allowed to move is the cheapest way to avoid warping. At 480p and 2 credits per second this is a very cheap draft before a costlier model.
The cheapest way to draft motion: 2 credits per second at 480p, 4 at 720p, any duration from 1 to 15 seconds with no minimum. Use it to test the camera move before paying for the final.
Long single takes of 5 to 30 seconds at 480p or 720p, 35 or 79 credits per second with a 5-second minimum, and reference video support billed on input plus output seconds.
Fixed 2K output with native stereo audio, 4 to 15 seconds at 46 credits per second, from text or a single first frame.
Fixed per-clip pricing at 720p, 1080p or 4K for 4, 6, 8 or 10 seconds, with up to 7 reference images or a first/last frame pair. 720p and 1080p cost the same.
A fixed 8-second clip at 38, 47 or 113 credits for 720p, 1080p or 4K, when you want one predictable price rather than a per-second meter.
Fast Fashion Dress-Up Taxi
Pop-up Book Paper Metropolis
Rainy Night Coffee Confrontation
Contrast in Style and Time
Woman Hiding Tears with Shirt
Candid Fashion Mirror Selfie
Celestial Palace of Immense Scale
AI Art Model Comparison
Cyberpunk Samurai in Neon Rain
Sweet Moment Shared
AI Hanfu Fashion Show
Hanok Blueprint Cinematic Visualization
Giant Hands Style Barbie Doll
First-Person Action POV
Teddy Bear's Emotional Release
Korean Shampoo Commercial
Bride's Internal Conflict
Fitness Journey AI Video
Q-version sticker character fails facial recognition
AI Ancient Chinese Fashion Show
10-Second Outfit Transformation
Woman's Weekly Fashion Showcase
Food Summon Freeze Pose
Tokyo Morning Vlog AdventureMost models meter per second, so duration multiplies the price. Draft on a cheap per-second model, then re-run the approved prompt once on the delivery model. The generate button shows the exact cost before each run.
Reduce what is allowed to move. Name the moving elements, keep the camera to a single move, and shorten the clip; long takes accumulate drift.
Use one explicit move per prompt, such as slow dolly forward or locked-off static, and state "no cuts". Describing several moves in one prompt invites the model to blend them.
Do not generate titles. Ask for a clean plate with empty space and add type in your editor, where it stays sharp and correct on every frame.
With a reference video, the Seedance family bills on input plus output seconds. Trim the reference to the part you actually need before uploading; the real duration is read automatically and shown before you generate.
Most video models bill per second at the chosen resolution, and a few bill a fixed price per clip. Grok Imagine Video 1.5 is 2 credits per second at 480p, Seedance 2.5 is 35 at 480p and 79 at 720p, MiniMax H3 is 46 at 2K, and Veo 3.1 Fast is a flat 38 to 113 credits for a fixed 8-second clip. The cost is shown on the generate button before you run.
Any model with first-frame or reference-image input. MiniMax H3 takes a single first frame, Gemini Omni Flash 1.1 takes up to 7 reference images or a first/last frame pair, and the Seedance 2.0 family takes up to 9 reference images. Starting from an approved still keeps the composition you already signed off.
Draft on Grok Imagine Video 1.5 at 480p for 2 credits per second, judge the camera move and the motion, then re-run the approved prompt once on the delivery model at the final resolution.
Yes. Topaz upscaling costs 10 credits per second at 1x or 2x and 18 at 4x, billed on the input duration for a single file up to 50MB, and the reverse-script tool turns an existing video back into a written shot list.
It depends on the model: Seedance 2.5 runs 5 to 30 seconds with a 5-second minimum, the Seedance 2.0 family and MiniMax H3 run 4 to 15 seconds, Grok Imagine Video 1.5 runs 1 to 15 seconds, and Veo 3.1 is fixed at 8 seconds.